Introduction: what the rainfall split actually means
When a rain gauge in an open clearing reports ten millimetres, it has not measured ten millimetres reaching forest soil. Some water passes through crown openings, some lands on leaves and branches and drips later, some follows bark toward trunks, and some remains temporarily on surfaces before evaporating. The practical task in a forest throughfall and stemflow study is to translate those pathways into one defensible water balance. The three main quantities are gross precipitation, throughfall and stemflow. Interception loss is usually estimated as their difference over a sufficiently well-defined period. This guide explains the mechanism, field measurements, conversion of collector volumes into depth, uncertainty, worked examples and interpretation without turning a residual into a claim of a directly observed evaporation flux.
The distinction matters to hydrologists, foresters, ecologists, watershed managers and students reading technical papers. A published percentage of rainfall interception might be a plot-average estimate for one wet season rather than a universal property of a tree species. A very wet storm and a sequence of tiny showers can produce different fractions even in the same stand. Nor is a collector below one picturesque tree a measurement of the whole forest. The aim is to ask what was actually sampled, what denominator was used, how the rainfall events were separated and whether the component depths refer to the same land area.
The four quantities and one conservation equation
Gross precipitation, often denoted P, is the rainfall depth measured above or outside the canopy that represents the precipitation arriving at the study stand. Throughfall, TF, is precipitation that passes through canopy openings or drips from vegetation onto the forest floor. Stemflow, SF, is the water routed along stems and trunks. An event interception residual, I, is commonly calculated as I = P − TF − SF when the terms are all expressed as equivalent depths over the same horizontal ground area. This is a bookkeeping relation, not a device measuring evaporation directly. It hides potential changes in temporary canopy storage, measurement errors and other relevant transfers.
One millimetre of water depth means one litre of water spread across one square metre. That equivalence makes rainfall gauges and collection volumes commensurable. A collection vessel measuring 2.4 litres beneath a funnel with a horizontal opening of 0.20 square metres has caught the equivalent of 12 millimetres over its own opening. That is a local throughfall observation; it is not yet the stand-average throughfall. For stemflow, the volume down a trunk is instead assigned to a defined plot ground area, or individual-tree crown or other explicitly chosen denominator. Using trunk surface area as the denominator changes the statistic and must never be silently mixed with plot-depth units.
A physical journey through the canopy
At the beginning of a dry-canopy storm, incoming droplets encounter leaves, needles, twigs, bark and epiphytes that can retain water. Some rain passes unobstructed through gaps. Small droplets may stick, merge and eventually drip after storage approaches an event-specific limit. Water reaching branch junctions may follow bark paths and collect at the stem base, but branch architecture also sends water away from the trunk. Wind can dislodge droplets, tilt rainfall paths or redistribute splash. During a break in the rain, a wet canopy may continue dripping while it also evaporates water. Thus each component can have a lag relative to the open rainfall record.
This is why a time series is richer than a storm-total pie chart. Throughfall can begin almost immediately in a crown gap, while a collector below dense foliage responds later. Stemflow may show a delayed start and then continue after rainfall stops. If a study ends a rainfall event at the last open-gauge tip, its downstream collectors may still be receiving drainage. The event boundary must allow for physically plausible delayed water, or describe how postponed drainage was allocated between events. The canopy is a temporary reservoir, not a filter that permanently removes a fixed percentage of every shower.
Define the spatial footprint before placing instruments
An appropriate study begins with a map showing stand boundaries, species composition, canopy cover, slope, access, plot size and open-gauge locations. The researcher must state whether the question concerns a particular tree, a planted row, an entire hectare, a forest age class or a watershed. These are different populations. Collectors in a sheltered stand interior can miss the edge behaviour of a fragmented forest, while a gauge close to buildings may measure neither the clearing precipitation nor the forest conditions accurately. The sampling unit is not simply the instrument: it may be a position, tree, plot or paired rainfall event.
Stratification can be useful when the canopy contains distinct crown classes, gaps and dense clusters. A sensible protocol decides before data collection whether sampling points are fixed, relocated randomly, or systematically moved between events. Moving a collector is not automatically better; it trades spatial replication against repeated measurement of the same microsite. In a sparse sample, one branch drip point can dominate the mean. State how collectors were selected rather than implying that a convenient installation is a random representation of the forest.
Gross rainfall measurement and exposure
The open-site gauge supplies the denominator in the interception calculation. Its position should avoid obvious overhead obstructions and undue turbulence without placing it so far from the study that the storm field becomes materially different. Record the collector orifice, mounting height, clock settings, measuring principle and maintenance history. A tipping-bucket gauge converts discrete tips to depth but can undercount during intense precipitation because water passes while the mechanism moves. A weighing gauge may resolve continuous accumulation but still faces wind-induced catch bias. Neither can be treated as an error-free reference.
For each storm, check whether the rainfall total comes from one gauge or a corrected combination. A forest-to-clearing comparison is particularly vulnerable when the clearing gauge and canopy collectors differ in wind exposure. A low apparent interception fraction might partly reflect an open gauge that undercaught wind-driven rain. Conversely, gross rainfall overestimation could inflate the residual. Technical translation should preserve words such as ‘measured’, ‘adjusted’ and ‘estimated’ because they indicate different evidential status. Retaining raw readings alongside any corrected series makes the calculation reproducible.
Throughfall collectors: volume, area and placement
A throughfall collector usually presents a defined horizontal aperture to falling or dripping water and conveys it into a protected container. Funnel diameter, opening area, rim level, splashing protection and overflow capacity matter. Under-canopy drops may be larger than open rainfall drops and can splash out of shallow funnels. Leaves, insects and debris can block entry, while a tilted aperture changes the effective horizontal capture area. A collector that floods before the next reading does not supply an ordinary total; its value should be flagged as a lower bound or missing under a stated rule.
Each collector is a local sample of a spatially variable field. Imagine six collectors returning 4, 7, 9, 11, 15 and 20 millimetres during one storm. Their arithmetic mean is 11 millimetres, but the range reveals strong patchiness. A report that prints only 11 millimetres hides the site-to-site dispersion that determines sampling uncertainty. If collectors have unequal intended plot weights, a weighted average may be needed instead. The scientific decision is how the set of apertures represents ground area, not whether a spreadsheet can calculate an average.
Stemflow collection and the tree-area denominator
Stemflow is commonly routed through a collar or spiral gutter around the trunk into a storage vessel or logger. The seal must guide water rather than leak beneath the collar, and the device must not send water artificially into the system from outside its intended bark surface. Rough bark, branching, trunk lean and stem enlargement all complicate installation. Measure the trees’ relevant attributes and explain the inclusion rule. A plot study might sample every trunk above a specified diameter or a stratified subset that is scaled to a tree census.
Suppose four fully monitored trees in a twenty-square-metre plot yield 1, 2, 0.5 and 1.5 litres of stemflow. Their sum is 5 litres. The plot-depth equivalent is 5 divided by 20, or 0.25 litres per square metre, hence 0.25 millimetres. Reporting the average of the four tree volumes as 1.25 litres does not itself tell a hydrologist how much stemflow reached each square metre of plot. The denominator must be recovered. When scaled from a sample, explain whether expansion weights come from crown area, trunk count, basal area or species classes, and do not conceal the assumption that sampled trees represent unsampled trees.
Reading an event time window
An event may be defined by a minimum rain-free interval, but there is no universal interval suitable for all canopy structures and climates. Storms that arrive closely together can begin on a wet canopy, which changes the amount of new storage needed. Collectors measured manually at fixed morning times may combine several showers. In a high-frequency record, lagged dripping can be attributed to the preceding storm if the rule is explicit, although uncertainty grows when rain resumes during drainage. This is a measurement-design issue, not a minor data-processing preference.
Before comparing event percentages, look for the start and end times of the open rainfall, throughfall and stemflow records. Are all instruments synchronized? Was daylight saving or a data logger timezone handled correctly? Did stemflow bottles remain untouched until late drainage finished? Was the previous event dry enough to reset canopy storage? If not, a high throughfall-to-precipitation ratio may reflect release of water stored during an earlier shower. The algebra can still be performed, but the physical attribution is less simple than the formula suggests.
Worked example one: a complete plot water balance
Consider a clearly hypothetical thirty-square-metre forest plot with an open rainfall depth P of 18.0 millimetres during an event. Twelve equal-area throughfall collectors report a plot-representative mean TF of 12.6 millimetres after valid-quality control. All sampled and scaled trees together send 27 litres of stemflow to their collars. Because one millimetre over one square metre is one litre, the plot-equivalent stemflow depth is SF = 27 / 30 = 0.90 millimetres. The interception residual becomes I = 18.0 − 12.6 − 0.90 = 4.50 millimetres.
The throughfall fraction is 12.6 / 18.0 = 0.700, or 70.0 percent. The stemflow fraction is 0.90 / 18.0 = 0.050, or 5.0 percent. The interception fraction is 4.50 / 18.0 = 0.250, or 25.0 percent. Their sum is exactly 100 percent in this arithmetic illustration, which is a useful closure check. It would be incorrect to say the instrument recorded 4.50 millimetres of evaporation; the value was calculated by subtraction and inherits uncertainty from each input. It would also be incorrect to divide the 27 litres by the sum of trunk cross-sectional areas unless the intention was to report a separate stemflow yield metric.
Worked example two: unequal throughfall-area weights
Imagine a plot divided into two recognized canopy classes. Dense crown area occupies 60 percent of the plot and its representative collectors average 7.0 millimetres for an event. Sparse crown area occupies 40 percent and averages 13.0 millimetres. The area-weighted throughfall is TF = (0.60 × 7.0) + (0.40 × 13.0) = 9.4 millimetres. A simple average of the class means gives 10.0 millimetres, overestimating plot throughfall by 0.6 millimetres because it treats the classes as equally extensive. The difference can be significant when precipitation is small.
If gross rainfall is 14.0 millimetres and stemflow is 0.40 millimetres over the plot, the consistent residual is 14.0 − 9.4 − 0.4 = 4.2 millimetres. Its fraction is 30 percent. With the unweighted throughfall value, the apparent interception would be 3.6 millimetres, or about 25.7 percent. The incorrect method would change the interpretation without a single sensor malfunction. The lesson is that representativeness and denominators precede precision. A reader should ask whether reported ‘average throughfall’ is the average of instruments, plots, events, canopy zones or spatial strata.
Worked example three: negative interception residual
Suppose the gross gauge reports 6.0 millimetres, throughfall collectors average 6.4 millimetres, and plot-scaled stemflow is 0.2 millimetres. The computed residual is −0.6 millimetres. Negative interception does not establish that the canopy manufactured water. It says that the simple event-accounting inputs do not close under the adopted boundaries and uncertainty. Possible explanations include delayed drainage from a previous shower, spatial mismatch between the open gauge and under-canopy catches, gauge undercatch, throughfall drip concentration at too few sample locations, or instrument and timing faults.
Do not automatically replace negative values with zero and then calculate annual totals as though nothing happened. Such censoring changes the aggregate estimate. Instead record the event, the raw data, the quality flags and the best explanation. If the analysis requires a physically constrained model, report modelled estimates separately from observed components and identify the constraint. A negative residual is often analytically useful because it forces researchers to inspect the assumptions that were otherwise invisible. Decide whether to retain, flag or omit it using a rule established before inspecting desired outcomes.
Throughfall coefficient and the problem of averaging ratios
The ratio TF/P is sometimes called a throughfall fraction or throughfall coefficient. Its interpretation depends on event scale. If a small storm of 2 millimetres yields 0.5 millimetres of throughfall, TF/P is 25 percent. If a large storm of 30 millimetres yields 26 millimetres, TF/P is about 86.7 percent. The mean of those two event fractions is 55.8 percent, but the total throughfall divided by total rainfall is 26.5 / 32 = 82.8 percent. Both numbers can be mathematically described, yet they answer different questions.
The unweighted mean tells us about the typical proportion across the two chosen events, giving a tiny shower the same influence as a large storm. The ratio of totals tells us what proportion of the entire observed rainfall volume appeared as throughfall. For an annual water budget, the second is generally the relevant water-volume ratio. A careful report can present both, with their names and denominators clearly labelled. Do not call an unweighted average of event percentages an annual rainfall partition without specifying the estimand.
Stemflow funneling ratio is not a plot-depth percentage
Some papers report a stemflow funneling ratio, a measure of how strongly the catchment of a tree concentrates rain toward its base relative to direct rainfall on the trunk basal area. This statistic can be much larger than one because an extensive crown and branching system channels water to a small trunk footprint. It is conceptually different from the percentage of plot rainfall conveyed as stemflow. Readers may mistakenly interpret a funneling ratio of ten as ten percent stemflow, even though ten is dimensionless and could mean much stronger concentration relative to the chosen reference area.
To translate it, write the numerator and denominator in full words before interpreting the number. If the numerator is stemflow volume and the denominator is gross rainfall depth times trunk basal area, the ratio indicates a reference-area concentration. It does not describe the fraction of rainfall over an entire stand without a separate scaling calculation. If the denominator is projected crown area instead, the value changes. Different publications can also use competing conventions. A figure caption should state the definition, rather than expecting readers to recognize an abbreviation whose meaning shifts across research groups.
Interception loss versus canopy storage change
The event residual I = P − TF − SF is often treated as interception loss, but the exact water balance for the canopy also contains changes in retained water storage and possibly other transfers according to the system boundary. If a canopy begins dry and ends dry, the net change in canopy storage can be close to zero, leaving evaporation of intercepted water as the dominant residual pathway under ordinary assumptions. If the event ends while branches are still wet, the canopy can be storing part of the missing water. If it begins wet from a previous event, the sign and allocation become more complicated.
This distinction matters especially in short observation windows and persistently humid environments. Wetness does not vanish just because rainfall stops; dripping can persist as retained droplets merge, and evaporation depends on available energy and atmospheric conditions. Authors should use phrases such as ‘estimated interception loss over the complete event and subsequent drainage period’ rather than claiming that all missing water evaporated instantly during the rainfall. If evaporation itself is modelled, specify the model structure and inputs, not merely the residual subtraction.
Why canopy structure changes the result
Needle orientation, broad-leaf shape, branch angles, bark roughness and crown architecture determine where droplets can be retained or routed. Leaf area and leaf angle influence interception opportunity, but high optical leaf area index alone cannot predict a fixed interception loss without rainfall characteristics, wetness history and aerodynamic conditions. A tall conifer stand and a deciduous broadleaf stand may have different stemflow fractions, yet a species name by itself is insufficient to quantify event partitioning. Trees also vary through seasons: leaf-out, senescence and pruning change effective collecting surfaces.
When comparing forests, look for stand age, canopy gap structure, basal area, tree density, leaf-on versus leaf-off periods and the distribution of rain-event sizes. At minimum, a good comparison should not confuse seasonal precipitation differences with intrinsic vegetation effects. If one site experienced mainly small showers and another mainly intense storms, their interception fractions need not differ because of the canopy alone. The scientific translation is conditional: this canopy, under these weather conditions, observed by these instruments, generated this measured partition.
Storm size and dry antecedent period
A small shower falling on a dry canopy can be largely retained before any substantial drainage begins. After the canopy is wetted, additional rainfall is more likely to become throughfall or stemflow. This explains a common pattern: the interception fraction declines with increasing storm depth. It is a tendency, not a universal deterministic curve. High winds may shake off droplets, intense rain may run off rapidly, and between-event evaporation can partly dry the canopy. The length of the preceding rain-free period is therefore a meaningful explanatory variable.
An empirical interception model may include storm depth, rain duration, mean or peak intensity, antecedent dry period, wind and vapour-pressure conditions. Before applying such a model, check which variables are predictors and whether their units match the original calibration. A logarithmic or saturating curve should not be extrapolated far outside the observed range. If a model predicts negative interception for an extreme input, the failure should lead to domain checks, not a fictional explanation that the canopy creates rainfall.
Rain intensity and sampling resolution
A total of twenty millimetres delivered in twenty minutes is physically different from twenty millimetres delivered across an entire day. Intense rain can overwhelm storage and collection systems, while prolonged gentle rainfall gives more opportunity for evaporation during the event. Throughfall can be concentrated in bursts when water at a branch junction reaches a drainage threshold. Ten-minute data can expose that timing, whereas manual daily bottles collapse the sequence into one cumulative amount. Neither resolution is automatically superior; it must match the question.
Report sampling frequency, logging interval and how missing records were handled. If a bucket tips at fixed increments, the time stamp denotes the moment of a tip, not necessarily the start of rainfall during the preceding interval. If the throughfall collector was emptied once daily, assigning its entire volume to one of two nighttime showers is unjustified. When converting an event depth into intensity, divide by the correct elapsed hours rather than by the number of calendar dates touched. Storm intensity should not be inferred from a daily total without finer temporal information.
Uncertainty propagation with units
A residual calculated from three uncertain numbers is itself uncertain. Suppose gross rainfall is 18.0 ± 0.5 millimetres, mean throughfall is 12.6 ± 0.8 millimetres and stemflow is 0.9 ± 0.2 millimetres, with the numbers representing standard uncertainties that are reasonably independent. The residual is 4.5 millimetres. A simple first-order quadrature gives uncertainty sqrt(0.5 squared + 0.8 squared + 0.2 squared), or about 0.96 millimetres. That calculation assumes independence and suitable uncertainty definitions; it is not automatically valid for correlated gauges or systematic exposure bias.
Do not casually turn this into ’25 ± 5 percent’ without propagating the shared rainfall denominator and its covariance with the residual. For a ratio R = I/P, a more careful approach propagates the underlying input vector or uses resampling that preserves event pairing. Systematic biases deserve separate treatment because adding more events does not make a miscalibrated gauge unbiased. If throughfall collectors are spatially clustered, their errors are not independent simply because twelve funnels exist. State whether the reported variation represents sensor accuracy, between-collector spatial spread, between-event variation, or uncertainty of the estimated mean.
Plot sampling, pseudoreplication and confidence intervals
Twelve funnels in one forest plot are spatial subsamples, not twelve independent forests. They help estimate mean throughfall in that plot but cannot independently establish how all forests of the same species behave. Equally, collecting a hundred storms from one season does not automatically represent the full climate distribution. The number of independent units depends on what claim is being made. An estimate of plot-specific annual partition may use events as repeated observations, but extrapolation to a region requires plots or sites with appropriate selection.
For throughfall variability, calculate the distribution across positions and consider spatial clustering. A confidence interval on the spatial mean should reflect actual sampling design. Bootstrap methods can be useful when resampling coherent spatial units or whole events, but a naive bootstrap of every individual time stamp creates artificial replication. If collectors are relocated between events, a different hierarchical structure arises. The report should distinguish spatial uncertainty from temporal uncertainty rather than supplying one attractive decimal and calling it a universal confidence interval.
Field quality control and failure logs
Each observation benefits from a simple record: instrument identifier, location, aperture area, event window, observed volume or tips, quality flag and any corrective action. Throughfall bottles can contain leaves, ants or spilled water; collars can detach; tubing can clog; data loggers can reset; open gauges can experience wind effects. A photograph of a badly blocked funnel is useful documentation, but it does not recover the lost precipitation. Flagging such an event preserves honesty and allows sensitivity analysis. Unrecorded substitutions create false precision.
Quality-control rules should be specified before final plots are inspected. For example, decide how to treat visible overflow, zero readings during a large storm, impossible negative cumulative volumes and outlying drip-point readings. Outliers are not necessarily faults: a collector under a drip point can legitimately exceed open rainfall locally. Remove a measurement only for a defensible physical or operational reason. Carry both raw and adjusted data, and show how excluding failed observations changes the estimated mean and interception residual.
What throughfall greater than gross rainfall means
A particular under-canopy funnel can collect more water than its corresponding open rainfall depth. Rain intercepted by a larger canopy surface may be concentrated into a smaller drip zone. One collector registering 28 millimetres when the open gauge reads 20 millimetres does not violate conservation at forest scale. It may indicate a focused drainage pathway. The problem arises when a handful of local peaks are mistaken for a representative ground-area average. The spatial design and weights determine whether the mean is defensible.
By contrast, an entire plot-mean throughfall exceeding gross precipitation for many well-separated dry-to-dry storms demands careful scrutiny. It may arise from geographic precipitation mismatch, insufficient sampling, gauge bias or inaccurate weights. When both TF and SF exceed P, the simple residual becomes negative. Researchers should report such cases transparently and diagnose the boundary conditions before describing them as a novel hydrologic process. Translating a figure correctly means separating local concentration from the regional or plot mass balance.
How to compare two forests fairly
Suppose Forest A reports 32 percent interception over 100 millimetres of gross rainfall and Forest B reports 18 percent over 400 millimetres. You cannot conclude from those figures alone that Forest A is intrinsically more effective at intercepting rain. Perhaps A experienced many small events after dry periods, while B was measured during long, intense storms. Perhaps stemflow was ignored at one site but measured at the other. Or one study sampled a dry-season canopy and the other an unusually wet season.
A fair comparison first aligns definitions: event duration, rainfall reference, inclusion of stemflow, plot area, seasonal window, event-size distribution and quality exclusions. Next inspect absolute depths as well as fractions: 32 percent of 100 millimetres is 32 millimetres, while 18 percent of 400 is 72 millimetres. The site with the lower percentage can have the larger total interception depth. If the research question concerns water available to soil, the absolute throughfall-plus-stemflow depth can be more relevant than the interception percentage alone.
Worked example four: monthly aggregation
Take three hypothetical events with gross rainfall depths of 5, 15 and 40 millimetres. Their respective throughfall depths are 1, 10 and 35 millimetres; plot-equivalent stemflow depths are 0.1, 0.5 and 1.0 millimetres. The residuals are 3.9, 4.5 and 4.0 millimetres. Their event interception fractions are 78 percent, 30 percent and 10 percent. The simple mean of those percentages is about 39.3 percent, an answer to the question ‘What was the average event fraction?’.
For the monthly water balance, gross precipitation totals 60 millimetres, throughfall totals 46 millimetres and stemflow totals 1.6 millimetres. The residual is 12.4 millimetres, or 20.7 percent of monthly precipitation. This is the relevant monthly rainfall-volume fraction. A monthly bar chart should show the 60, 46, 1.6 and 12.4 millimetres on a consistent basis. If the events were not the only rainfall that month, the label ‘monthly’ is wrong; call it ‘sum of three selected events’. Completeness is part of the denominator.
Stemflow concentration at the tree base
Stemflow may deliver a relatively small plot-wide depth while creating a locally large infiltration input near the trunk. Five litres directed into a narrow collar of soil can produce conditions different from five litres evenly distributed over a large plot. This local hydrology is why a small stemflow percentage can still matter to soil chemistry and root-zone heterogeneity. But it does not imply that all stemflow infiltrates immediately or stays near the base: soil surface slope, macropores, litter, runoff and stem-root architecture influence movement afterward.
When interpreting soil wetting near trees, do not automatically attribute a pulse to stemflow simply because a nearby sensor responds to rain. Compare timing, sensor depth, topography and possible lateral redistribution. A stemflow collector estimates water following the trunk; a soil probe estimates local water content or potential. They measure different quantities and require an explicit transfer model to link them. This distinction preserves the meaning of each instrument and prevents a plausible mechanism from being presented as a directly demonstrated one.
Interception in drought and dry periods
A forest can retain a notable fraction of small rain events, reducing direct precipitation delivery to the floor in that particular event. Yet describing interception as universally harmful to drought resilience is too simple. Canopies also shade the soil, influence aerodynamic exchange, alter snow deposition where relevant and affect transpiration and root uptake. A full ecosystem water balance is broader than rain partition. The event residual tells us how incoming rainfall was routed, not whether the forest’s net influence on seasonal water availability is positive or negative.
Dry-canopy storage capacity and antecedent wetness should therefore be reported separately from root-zone soil water. In an irrigation or forest management decision, ask what outcome matters: immediate surface wetting, recharge, soil moisture persistence, tree water use or downstream streamflow. The interception equation informs only part of that chain. It cannot alone quantify annual groundwater recharge or establish plant water stress. Linking measurements is valuable when each additional step is supported by independent observations or an identified model.
The difference between laboratory, field and modeled interception
A laboratory leaf wetting experiment can reveal local retention under controlled spray, but its spray droplet spectrum, wind and foliage orientation may not resemble forest rain. A field plot observes real weather but has spatial sampling and instrument exposure problems. A mechanistic model may represent canopy storage, drainage and evaporation in detail while depending on calibrated parameters. These sources of evidence are complementary rather than interchangeable. When a paper compares them, translate which quantities are directly observed and which are simulated or inferred.
Models may include an interception storage threshold and drainage function, while empirical relationships relate residual depth to storm size. Before taking parameters from one environment to another, ask whether canopy architecture, precipitation regime and measurement definitions match. A fitted parameter may absorb unmodelled bias and need not correspond to a literal physical capacity. Good reporting identifies calibration and validation data separately, avoids evaluating a model solely on the storms used to fit it, and shows uncertainty or failure cases.
Practical data table and reproducible steps
A transparent dataset can have one row per event with an event ID, start and end time, open rainfall P, each throughfall collector depth, mean and spatial spread of throughfall, each tree’s stemflow volume, plot-scaled SF, computed interception residual and data-quality notes. Keep collector metadata in a separate linked table: aperture areas, exact positions, maintenance dates and tree characteristics. This separation prevents accidental overwriting of a collector’s calibration when the field record is updated.
To reproduce the analysis, first verify units and date boundaries. Convert collection volumes to appropriate depths; combine spatial samples with the declared weights; scale stemflow to the plot area; subtract to obtain the residual; inspect closure and missing values; then summarize by event, season and complete study interval using clearly named denominators. The plot should not silently substitute a daily minimum for a missing reading or assume that an empty bottle means zero flow when it was actually unvisited. A reader who can repeat the workflow has gained much more than a memorized formula.
How to read figures and maps critically
A scatterplot of event interception residual against gross rainfall may show a saturating absolute response or a declining fractional response, but check what is on the y-axis. Interception depth in millimetres and interception percentage are not interchangeable. A plot of TF/P against P can show mathematical coupling because P appears in both axes, so apparent correlations need physical context. For a time-series graph, note whether throughfall is plotted as cumulative depth, rate or collector-average volume. Similar-shaped curves can have incompatible units.
Maps of throughfall spatial variability should identify the interpolation method and number of actual sampling positions. Smooth coloured surfaces between eight collectors may be visually appealing but are not direct observations at every pixel. A confidence band based on event variability does not necessarily show spatial sampling uncertainty. Ask whether figure annotations distinguish observed data, modelled surfaces and extrapolated values. If a chart groups wet-season and dry-season storms, look at event-size distributions before calling the difference a seasonal canopy effect.
Common errors and repairs
One error is to subtract stemflow litres directly from rainfall millimetres. Repair it by dividing stemflow volume by the correct plot area first. Another is to average collector volumes when funnels differ in aperture size. Convert each volume to a depth using that collector’s own horizontal opening before combining depths. A third error is treating the local maximum drip-point reading as an estimate of stand mean; repair it with adequate spatial sampling or clearly label it local. A fourth is reporting daily mean interception percentages as the month’s water-volume fraction. Sum component depths first, then divide by total gross rainfall for the volumetric proportion.
Other failures include omitting stemflow but labelling the residual total canopy interception, conflating open rainfall with nearest weather-station precipitation, censoring negative residuals without disclosure, using incomplete event windows, and assigning tube overflow to a valid zero. A useful repair is to have a second reader reconstruct the calculation from raw event rows. Ask them to identify the system boundary, units and denominator of every number. If they cannot, the reporting needs clarification even if the arithmetic is correct.
Independent transfer task one with answer
A twenty-five-square-metre plot records 12 millimetres of gross rainfall. The plot-weighted throughfall mean is 8.4 millimetres. Stemflow collectors hold 20 litres from all relevant trunks. Calculate the plot-equivalent stemflow depth, the interception residual and the corresponding fraction of gross rainfall. Pause before reading the answer and decide whether the stemflow volume should be divided by the collector opening, the trunk basal area or the plot area.
The correct plot-equivalent stemflow is 20 / 25 = 0.8 millimetres. The interception residual is 12 − 8.4 − 0.8 = 2.8 millimetres, and its fraction is 2.8 / 12 = 0.2333, or about 23.3 percent. The three plot fractions are 70 percent throughfall, 6.7 percent stemflow and 23.3 percent interception. They sum to approximately 100 percent after rounding. A claim that evaporation was directly measured would still be unwarranted: the residual derives from the three measured or aggregated inputs.
Independent transfer task two with answer
Two equal-length observation weeks record different rainfall totals. Week A has P = 8 millimetres and interception residual 4 millimetres. Week B has P = 72 millimetres and interception residual 9 millimetres. What percentage of incoming precipitation was intercepted across both weeks, and why is the mean of the two weekly percentages not the same number? This task tests whether you can preserve a water-volume denominator rather than averaging visually convenient ratios.
The two interception depths sum to 13 millimetres and gross rainfall totals 80 millimetres. The combined fraction is 13 / 80 = 16.25 percent. Week A’s fraction is 50 percent, while Week B’s is 12.5 percent; their unweighted mean is 31.25 percent. The combined fraction gives rainfall in Week B nine times as much weight because nine times as much water arrived. Neither statistic is an arithmetic mistake. The mistake is labelling one with the interpretation of the other. State explicitly whether the claim concerns an average observation period or total water partition.
Independent transfer task three with answer
A paper gives gross rainfall of 10 millimetres, one throughfall collector with 1.2 litres caught by a 0.10-square-metre opening, and another collector with 1.5 litres caught by a 0.30-square-metre opening. What local depths do the instruments represent? Should the two captured volumes simply be averaged? If the two collectors are equally representative of ground area and other sampling assumptions hold, what is their simple mean depth?
The first local depth is 1.2 / 0.10 = 12 millimetres. The second is 1.5 / 0.30 = 5 millimetres. Their mean depth under the stated equal-representativeness assumption is 8.5 millimetres. Averaging the volumes as 1.35 litres would not yield a meaningful average depth, because the funnels had different apertures. The first collector exceeds gross rainfall locally, which is physically plausible beneath a focused drip point. Whether two instruments adequately represent a plot is a separate uncertainty question and cannot be settled by arithmetic alone.
Independent transfer task four with answer
A study labels its monthly event-average interception as 40 percent. Another table reports total monthly rainfall of 200 millimetres and a total monthly interception residual of 50 millimetres. Are the statements necessarily contradictory? What must be checked before accusing the authors of an error? Think about the chosen events, aggregation method and possible missing observations.
The ratio of total monthly interception to total monthly rainfall is 50 / 200 = 25 percent. An unweighted mean of individual event fractions can nevertheless equal 40 percent if small storms with high percentages are numerous. First check whether ’40 percent’ is the mean of event percentages while ’25 percent’ is the water-volume-weighted monthly ratio. Also check whether both results cover the same complete month, including events discarded for sensor failure. If both are described as the same monthly volume fraction for identical data, the inconsistency needs correction. If they represent different estimands, both may be legitimate with precise labels.
FAQ: what does ‘intercepted’ mean?
In everyday speech, a canopy ‘intercepts’ rain when leaves and branches catch it. In an event water-balance paper, ‘interception loss’ often names the residual after subtracting measured throughfall and stemflow from gross rainfall. The terms are related but not identical: water caught on a leaf may later drip and count as throughfall. Only the water not reaching the forest floor within the chosen system boundary and period contributes to the residual. Always read the operational definition in the methods.
Can stemflow exceed throughfall? In plot-depth terms it is often smaller, but there is no algebraic rule requiring this in every tree or every storm. Can the interception residual be negative? Yes, as a calculated result under imperfect boundaries and observations; interpret it as a closure and quality-control problem rather than water creation. Is gross rainfall the same as precipitation at a distant weather station? Not necessarily. Spatial rainfall differences and gauge exposure can introduce bias. These questions are simpler once the measurements and denominators are named explicitly.
FAQ: which summary should a report lead with?
For a plot water budget, present the total gross rainfall, total plot-equivalent throughfall, total plot-equivalent stemflow and their residual in millimetres, followed by percentages relative to the same gross rainfall total. For a storm-process study, event-by-event curves and rainfall intensity may deserve prominence. For a spatial study, provide the throughfall distribution across collectors, the actual sample count and the canopy position map. A single interception percentage is too compressed to serve all three purposes.
Should a field team use more collectors or sample more storms? That depends on whether uncertainty is mostly spatial or temporal. A small number of widely dispersed collectors can be inadequate for patchy throughfall even across many storms. A beautifully replicated spatial network operating only during one odd afternoon cannot establish annual behaviour. A balanced design plans both dimensions. Where resources are limited, state the scope of inference instead of using larger words to conceal a small dataset. The strongest method section tells readers exactly what the measurements can support.
Concluding method: translate the boundary before the number
The most useful habit in reading a forest rainfall-partition figure is to draw the system boundary first. Rain falls onto a defined stand. One share reaches the ground as throughfall, another travels down trunks as stemflow, and the remainder is a residual that can involve interception evaporation and temporary storage depending on the observation window. Convert every collector volume to a valid depth, aggregate it over the correct ground area, and pair all components over the same time interval. Only then calculate fractions.
The method travels beyond forests. Similar discipline is needed when translating evaporation, deposition, fluxes, nutrient budgets and any scientific chart that uses a mixture of local measurements and landscape estimates. Words such as ‘observed’, ‘scaled’, ‘weighted’, ‘modelled’ and ‘residual’ are not decoration: they are labels for the chain of evidence. Keeping them precise makes technical English readable without flattening the science.
Sources and verified related reading
US Forest Service: Rates and Implications of Rainfall Interception in a Coastal Redwood Forest
US Forest Service: Sampling Throughfall and Stemflow in Young Loblolly Pine Plantations
FAO: Introduction to Evapotranspiration
eduKate: Weighing Lysimeter Mass Changes and Actual Evapotranspiration